VCs themselves probably suffer from chronic overestimation of their own intelligence, but there just aren't many good signals at the stage of companies they're looking at. No customers, no revenue; often just an idea and hopefully a prototype. GitHub stars are as good of a signal as letters of intent, which is to say: a bad signal, but at least a signal. Other than that, they have to just evaluate what the founders are telling them (generally unrealistically optimistic at best) and whatever market research they can do (which is hard enough for the founders to do for their own product; image doing this for a dozen different companies every day).

Of course GitHub stars are a terrible signal, but the bar for signal quality is just really low.